1
Transitioning from Unintelligible
to Intelligent Documentation
Session ID# PE5, February 11, 2019
Peter Basch, MD, MACP; Senior Director, MedStar Health
Qammer A. Bokhari, MD, MBA, MHSA, CPHIMS; VP/CMIO, Advent Health
2
Peter Basch, MD, MACP
Has no real or apparent conflicts of interest to report.
Conflict of Interest
3
Qammer A. Bokhari MD, MBA, MHSA, CPHIMS
Interest or their Agents (e.g., speakers’ bureau): Personally invested
(evangelist) in simplifying documentation through AI enabled
technology
Ownership Interest (stocks, stock options or other ownership
interest excluding diversified mutual funds): Angel Investor in AI
enabled Speech Recognition Technology
Conflict of Interest
4
• EHRs and E&M Documentation Guidelines – A Tragedy in 5
Parts
• Artificial Intelligence, its evolution and potential to simplify
and improve documentation
• Potential to leverage changes to Documentation Guidelines
for 2021 that could improve EHR UI and functionality
• What should the role of the clinician be in documentation
post 2021?
• Questions
Agenda
5
• Discuss the impact of the E&M documentation guidelines
on medical documentation and EHR functionality
• Compare the newly modified E&M guidelines with the prior
guidelines
• Describe evolution of AI in documentation and how it can
be used to formulate cogent documentation
• Application of AI-assisted documentation
• Discuss the impact of E&M reform and Artificial Intelligence
on EHR usability and usefulness
Learning Objectives
6
• Part 1 – Documentation without regulations
• Part 2 – Enter the Documentation Guidelines
• Part 3 – Enter the EHR
• Part 4 – “I spend more time on my EHR than I do with patients”
• Part 5 – Enter CMS and ONC
EHRs and E&M Documentation Guidelines –
A Tragedy in 5 Parts
7
It (Was) a Wonderful Life
Image in Public Domain – downloaded from
Wikimedia
8
It (Was) a Wonderful Life (sometimes)
All images licensed for use in presentation from
Shutterstock.com
9
Enter Evaluation and Management
Documentation Guidelines
Image created by authors (RB, AS, and PB) for use in article
10
• Elegant narrative
persisted at its own
peril
• The emergence of
the hybrid note
Impact of Documentation Guidelines on
Paper Records
Image licensed for use in presentation from Shutterstock.com
11
Enter the EHR
Image licensed for use in presentation from Shutterstock.com
12
Coding Software vs. CDS
Images on the left created by PB, Image on the right licensed for use in presentation from Shutterstock.com
13
I Spend More Time on My EHR than I
Do on Patient Care
Image licensed for use in presentation from Shutterstock.com
14
CMS and ONC are Listening
Images in Public Domain – downloaded from HHS.gov
15
CHANGES
• Medical necessity – no longer
necessary to document medical
necessity for home visits
• Redundant documentation – may
choose to ONLY document changes
to history and exam, and/or ONLY
refer to lists
• Documentation permitted by
others (including the patient) – may
choose to use staff or patient entered
CC and history
• Duplication of documentation by
teaching attendings – no longer
required
POSSIBLE IMPACT
• Small – currently easily satisfied by
templated attestation
• None to Substantial – need
clarification from Medicare carriers as
to exactly what is permitted. EHRs
may not yet support full potential
• None to Moderate – organizational
policies and medical professionalism
may dictate against using this change
• None to Minimal – need clarification
from Medicare carriers as to exactly
what is permitted, and if this applies to
student documentation, organizational
policies and medical professionalism
may dictate against using this change
The 2019 Medicare Physician Fee Schedule:
Current Changes to Documentation Guidelines
16
Options include
• Continuing to document as you
are now
• E&M compliance ONLY at level 2
• E&M compliance ONLY for Level
2 MDM
• Time-based documentation
– Medical necessity for visit
– Time spent F2F
– Rest… up to you
Implications for You and Your EHR
• None
• Notes would likely be significantly
shorter, more relevant
• Notes would likely be different, and
significantly shorter, more relevant
• Notes would likely be different, and
significantly shorter, more relevant
– Most interesting potential for
how the “house of medicine”
could leverage this option
The 2019 Medicare Physician Fee Schedule:
Changes to Documentation Guidelines Proposed
for 2021 and Beyond (Level 2 – 4 Visits)
17
An Integrated Intelligent Decision Support System (IIDSS) with
Real Time Clinical and Financial Surveillance
The Vision
18
Respects Physician Practice Autonomy
Exception:
• Deviation from Care Pathways!
– Evidence based practice (proven methodology)
– Practice based evidence (real world experiences)
• About to Commit
– An error of Omission
– An error of Co-mission
• New Developments
– Regulatory / Mandates
– New Guidelines, advisories or recommendations
Point of Care Advanced/Intelligent
Decision Support
19
Clinical Rules &
Algorithms (CDSS)
Artificial
Intelligence
Predictive
Modeling
Intelligent Decision
Support System
(IDSS)
Surveillance
Engine(IDSS)
Dictated Physician Encounter Note
DATE: 12/29/2010 13:45
REASON FOR CONSULTATION: Acute myocardial infarction.
HISTORY OF PRESENT ILLNESS: The patient is a 51-year-old without significant past medical history on no
medication. He is a heavy smoker who comes to the Emergency Room with 2 days of chest pain. The
patient started to have pain sometime on Saturday during the day. It was in her chest radiating up to her
neck as it also hurt to breathe. This persisted for the next 2 days. She called her friend Monday morning,
brought her to the Emergency Room. She is complaining of ongoing chest pain which she feels is similar
to her presenting pain; however, it hurts to move or to take deep breaths as it goes up to her neck and
jaw. It is a little better sitting forward. She has not had any of this discomfort prior to the onset on
Saturday.
Her risk factors are she smokes at least 1 pack a day. She was young and question whether she has
hypertension, but she is not treated. She has no diabetes from looking in her record on SRS. She did have
an elevated LDL of 150 back in 2007 and is not on treatment and drinks at least moderate alcohol.
Her son and friend were with her when I examined the patient. She was clearly in some
distress and complaining of his discomfort. Difficult to get a good complete history since
the patient is in distress.
Her CK-MB and troponin I were 3173, 98.8 and 58.7, respectively,
BUN 23, creatinine 1.3, AST 607, ALT 53, alkaline phosphatase 130. Her white count 18.5, hemoglobin
15.4, hematocrit 45.9. Her MCV 108.6, increased absolute neutrophil count of 16%, normal INR and
electrocardiogram showed inferior myocardial infarction with ST depression of up to 2 mm, particularly
in V3, 4 and 5. Chest x-ray showed what appeared to be cardiomegaly without congestive heart failure.
On exam, her blood pressure was in 180/70, her pulse 104. Skin was warm and dry. She appeared in
some distress. Neck was supple. Carotid: No bruits. No jugular venous distention. Lungs were clear. She
had normal heart sounds with what appeared to be a gallop rhythm and a 2/6 systolic murmur at the
apex. Point of maximal impulse was somewhat displaced laterally. Abdomen was soft. Extremities, she
had good peripheral pulses, no cyanosis, clubbing, or edema.
A stat echocardiogram done showed a very extensive inferior, posterior and lateral areas of akinesis; her
anterior wall contracting normally. She had moderate mitral regurgitation, mild-to-moderate tricuspid
regurgitation with an elevated pulmonary artery pressure estimate probably around 50 and there was
no significant pericardial effusion.
ASSESSMENT AND PLAN: This is a 51-year-old who has had an extensive inferior posterior
lateral myocardial infarction and moderate mitral regurgitation as a consequence. She is not in heart
failure and apparently her myocardial infarction began on Saturday and is ongoing. Whether her pain is
now all infarct pericardotomy syndrome or ongoing ischemia is unclear. She says pain is the same
although there is a pleuritic component. She does have ongoing ischemic ST depression of up to 2 mm,
which could represent posterior infarct. At this point, I would proceed to cardiac catheterization and
recommendations will be pending the results.
Discharge Plan:
1) beta blocker c lopressor 50mg PO BID
2) Start Cardiac diet
3) Follow up 3 months
4) Lipid profile
Dictated by: Dr Cardiology, MD
Recommendation: Consider
changing diagnosis to “Healthcare
Associated Pneumonia (HAI)”
Reasoning: Previous history of
hospitalization in the past 3 weeks
Source: HIE
Accept
Cancel
Remin
d Later
Image downloaded from Public Domain
Legacy
Decision
Support
Smart
Decision
Support
Advance /
Intelligent
Decision Support
Artificial
Intelligence
Neural
Networks
(Network of AI’s)
Evolution of AI in Decision Support
Advisors presents
recommendations to aid
decision making
Primarily relies on
discrete data
Involves multiple
algorithms
Learning & reasoning
Presents precise
decision and reasoning
for an action
Real-time surveillance
of discrete & non-
discrete data
Involves complex
algorithms
Learning, reasoning, forecasting & answer “what
ifs” (runs simulations)
AI & NN implements decision automatically
Works independently and at times requires no
action from the Decision Maker (Autonomous)
Highly complex and intercommunicating algorithms
(AI & NN)
Relies on discrete data
Historical Evidence
Based
Simpler rules & alerts
Legacy
Decision
Support
Smart
Decision
Support
Advance /
Intelligent
Decision Support
Artificial
Intelligence
Neural
Networks
Intelligent Voice
Recognition
Computer Assisted
Physician Documentation
(CAPD)
Sepsis Advisors …
Evolution of AI in Decision Support
Confidential & Proprietary
VTE Prophylaxis
Radiology Advisors
Sepsis Alert …
Drug Interactions
Dose Range
Allergy Alerts …
… Oncology
… Radiology
… Pathology
Virtual Digital Assistants
Interactive Voice
Recognition
Dr. Watson…
LDS
SDS
IDS
NN
AI
Images downloaded from Public Domain
Traditional Cruise Control
Adaptive Cruise Control
Lane Departure & Blind
Spot Warnings & Assist
Semi Autonomous Driving
Autonomous Driving
Legacy
Decision
Support
Smart
Decision
Support
Advance /
Intelligent
Decision Support
Artificial
Intelligence
Neural
Networks
Evolution of Documentation
Traditional Transcription
Hand written Notes
Confidential & Proprietary
LDS
SDS
IDS
NN
AI
Transcription with
Interactive Prompts
Ambient Intelligence
Documentation by Exception
i.e. Intelligent note generation
by combining Physician
documented exceptions with
past notes patterns
Ambient Intelligence
Documentation as a by
Product of Patient / Doctor
conversation
Images downloaded from Public Domain
Realtime
Documentation
Using Front-end Voice
Recognition Tools or Virtual
Scribes
23
Away from Computer Care to Patient Care
23
Ambient
Intelligence
Augmented
Intelligence
Images downloaded from Public Domain
24
Application of NLP/NLU/Machine Learning
in Clinical Documentation
Note: NLP - Natural Language Parsing & NLU - Natural Language Understanding
Chart Abstraction
Quality Measure
Clinical Documentation
Improvement / Integrity
Case Management /
Working DRG
Determination
Speech Recognition
(Front-end Voice Recognition)
Image Recognition
(Diagnostics: Radiology / Pathology)
Image to Text – Clinical Reports
(Recognition & Extraction of
Clinical Concepts)
Medical Transcription
Virtual Scribes
Real-time Clinical
Documentation Advisors
Point of Care / Front-end
Back Office / Back-end
25
Speech Recognition - Rapid Adoption
Frontend Voice Recognition (FEVR)
0
500
1000
1500
2000
2500
3000
3500
2017-04 2017-05 2017-06 2017-07 2017-08 2017-09 2017-10 2017-11 2017-12 2018-01 2018-02 2018-03
User Counts
Active Users vs. Enabled Users
Active Users Enabled Users
AdventHealth
Adoption – 72% of enabled users
Wins
5 year goal of 4500 users, reached in 14 months
Avgerage Time Saved: 82 mins
Range: 40 mins-220 mins
Improved documentation quality - bring back “Narrative” & Patient Story
Enable opportunities for real-time intelligent decision support
CDI Achievements
36-69% increase in number of
charts reviewed with no staffing
increase
27% increase in number of
clarifications sent with no
staffing increase
36% increase in the clarification
rate
AI reduced waste by
highlighting the charts that had
opportunity for improved
accuracy and moved charts with
less opportunity to the bottom of
the list
AI Enabled CDI Workflow
AI Achievements
26
Emergence of Clinical Documentation
Advisors - Transition to Realtime Nudges
Realtime Clinical Documentation Improvement (CDI), Computer Assisted
Physician Documentation (CAPD), Hierarchal Condition Category (HCC)
27
From Alerting … to Nudging …
to Best Practices
Advice
Only
Advice
+
Actionable
Decision
Advice
+
Actionable
Decision
+
Reference
to
Best
Practice
• Enables MU, decision support and direct billing
• But can result in lower documentation quality
(overly structured templates, cut & paste,… )
• May negatively affect physician productivity,
patient detail and overall care
EMR Direct Data Entry
Structured and encoded information
Handwritten Documentation
Unstructured notes
• Short to the point
• Told the patient “story”
• Illegibility – huge issue
• Cannot be reused
The Documentation Journey
Narrative Documentation
Unstructured notes, AI Templates / Macros
• Very expressive – tells the patient “story”
• More meaningful & useful to clinicians
• Incorporates AI powered templates & macros
• Simple. efficient & effective
.
AdventHealth 3 Year Strategy – iConnect Hospitals
Digital Assistant
Enabled EMR
Focus on Patient Care,
Not Computer Care
Realtime
Intelligent Decision
Support
Documentation
created as a
By Product of Doctor
Patient interaction
> 2020
Real-time
Intelligent Decision
Support
Documentation by
Exception
Eliminate Transcription
(Where Can)
Enable Voice Ordering
2019 - 20
Real-time / Near-time
Documentation
Reduce Transcription
(Where Can)
Reduce Copy/Paste
Move to Narrative
Documentation
Transition PowerNotes
Providers to DYN DOC
(Where Can)
2018 -19
Transcription
standardization to
single vendor
(Outsourced &
Inhouse)
Deployed
Front-end
Documentation Tools
2016 - 17
29
30
• A dramatic reduction in copy-paste
• EHR presentation could evolve from
click-boxes to informational displays
– Longitudinal / time-line views
– Screen stare could change from
headache inducing distraction to
useful, educational, engaging
• Distinction between visit or “progress
note” and “all the news that’s fit to print”
• Enable simple, efficient and effective
documentation
• Assist in reducing burden of documentation
– Documentation by exception
– Documentation as a by product
• Enable real-time or near-time documentation
– Reduce / elimination after hours
documentation / chart completion
– Enabling real-time Intelligent clinical or
operational alerting / nudging
– Transition of back office support functions to
Point of care transactions; medical
transcription, chart abstraction, clinical
documentation improvement, computer
assisted coding …
Potential Impact to EHR from Leveraging E&M
Reform and AI
31
• When there are no prescriptive regulatory requirements
concerning documentation and there is an ability to auto-generate
a visit “transcript” - is there a role for the clinician in crafting
documentation?
What Should Ideal Documentation Look Like?
